siepic / siepic_ebeam_pdk Goto Github PK

.lsf and .spi files are saved in %KLAYOUT_HOME%/tmp. They should be deleted after INTERCONNECT gets closed.

Query sub-PCell for pin location, rather than re-calculate at the higher level. e.g,. location of routing waveguides should be matched to component, rather than manually calculated.

When two waveguides are overlappping and connected to the GC, we get an error:

> optical_components[k].pins[m]
list index out of range
  (eval):1
> optical_components[k].npins
2
> optical_components[k].component
'ebeam_gc_te1550'

This also happens when the GC is disconnected.

But when correctly connected to one waveguide, there is no error.

Of fabricated devices, so we can see the SEM vs. GDS, overlaid.

Y-branch snap to Array of GCs always snaps to the bottom GC.

For the Directional Coupler, which is already in the PDK. The DC has many parameters (g, w, r, length, wavelength range) hence providing a full parameter sweep is not reasable. The user may ask for a component that hasn't been simulated (e.g., w = 300 nm, g = 250 nm, L = 1.5 µm, r = 9.9 µm, lambda = 1310 nm). Need a method to get this filled in:

• in KLayout, user launches a simulation; INTERCONNECT issues a warning that the model doesn't exist and the simulation isn't accurate.
• INTERCONNECT launches FDTD; result is S-parameters
  • user registers the Sparam model into the CML (eventually automated):
  • KLayout launches INTERCONNECT, to find out the path of the EBeam CML. ?get("local path");
  • KLayout copies the Sparam file into the CML, and edits the XML file to add the new component
• INTERCONNECT now has the correct model.

For new components / user generated components, if the user only wants to simulate the SParams:

• use the above item, of launching FDTD to get Sparam.
• if someone wants to add this to the PDK, they would manually edit the CML directly.

Add cross-section capabilities, for advanced Active processes.

http://sourceforge.net/projects/xsectionklayout/

We have a CML for the regular DC, namely ebeam_dc_te1550, but we don't have a KLayout PCell.

Xu implemented the CML, so you can check with him for parameters, FDTD model, etc.
See page 9 of ebeam_Lumerical_CML_User_Guide_2015_08_24.pdf

Also add automated CML generation via FDTD.

Parameters:

• gap: gap
• radius: Radius of the bus. Assume the output is 2 conventional bends
• pitch: vertical pitch (separation + w) for the in/output waveguides.
• wg_width: Waveguide width
• coupling_length: coupling length
• polarization: 'TE' or 'TM'; for use in INTERCONNECT calling FDTD to simulate one of these if pol is defined, or simulate both if undefined.
• wavelength: default 1550. Consider a simulation bandwidth of 10% of the central wavelength. Used by the INTERCONNECT-FDTD simulation.

Annotation:

• kappa
• t
• excess loss

Perhaps add a layout property for adjustable radius.

SParameters for Ebeam crossing component, please.

None of the files in the verification section detects intersecting waveguides. @lukasc-ubc

It shrinks the length rather than increases it... Did the bug get re-introduced? I know you fixed this before.

• TM GC does not excite the TM mode.
• Y branch missing the TM simulations.

strip waveguide model correctly has TE and TM modes.

other components - what do they do for TM?

closed GUI, then tried to run it again:

Known issue:

In a hierarchical design, you need to complete the waveguides for the lower level cells first. If you run the script at the top level, the paths in the subcells get incorrectly converted.

Easiest fix:

Check that the path is in the present cell.

• if you draw a straight waveguide when connection = any angle, the Waveguide script makes a "bad" waveguide.

Verification, with 1000s components.
display run time.

Update all components to support TE, TM or both polarizations, for any centre wavelength, using the automated CML generation via FDTD.

Add to each component:

• pol: 'TE' or 'TM'; for use in INTERCONNECT calling FDTD to simulate one of these if pol is defined, or simulate both if undefined.
• wavelength: default 1550. Consider a simulation bandwidth of 10% of the central wavelength. Used by the INTERCONNECT-FDTD simulation.

rename components,

e.g., ebeam_y_1550 -> ebeam_y
e.g., ebeam_dc_te1550 -> ebeam_dc

A* in Python:

http://www.redblobgames.com/pathfinding/a-star/implementation.html

Theta*

http://aigamedev.com/open/tutorials/theta-star-any-angle-paths/

Grid with line movement: http://www.redblobgames.com/pathfinding/grids/algorithms.html

http://code.activestate.com/recipes/578919-python-a-pathfinding-with-binary-heap/

http://theory.stanford.edu/~amitp/GameProgramming/Heuristics.html

Need a thorough set of DRC rules.
Also avoid false errors like the grating coupler corners.

either leave as-is, or add functions to PCells:
http://klayout.de/forum/comments.php?DiscussionID=757&page=1#Item_8

The Path_to_Waveguide takes a single waveguide (<10) and connects both ends to the same pin. Thus wg_length = 0.

Check if waveguide ends are non-Manhattan, since they won't be able to connect to a pin.

The ordering of the nets in the netlist is arbitrary. Need to ensure that it is in a consistent order.

when running Snap, or verification.

Currently working on it.

• check that there is only 1 laser.
• search the Netlist and remove components that are not connected to a laser/detector. Performance improvement especially for MC simulation.

A request to implement a Layout PCell + CML for a fully-parameterized directional coupler for pulley-coupled resonators, with the following parameters:

• g: gap
• rr: Radius of the ring.
• rb: Radius of the bus (set the default to be 2Rr as in Fig 1).
• wr: Waveguide width for the ring
• wb: Waveguide width for the bus coupler
• rc: Radius of the coupler, for the waveguide leaving the ring.
• theta: The arc for the coupler
• pol: 'TE' or 'TM'; for use in INTERCONNECT calling FDTD to simulate one of these if pol is defined, or simulate both if undefined.
• wavelength: default 1550. Consider a simulation bandwidth of 10% of the central wavelength. Used by the INTERCONNECT-FDTD simulation.

name: ebeam_dc_halfring_pulley

generate an icon (SWG file)

See Figure 1 in the following paper:

Nourhan Eid, Hasitha Jayatilleka, Michael Caverley, Sudip Shekhar, Lukas Chrostowski, Nicolas A. Jaeger, "Wide FSR Silicon-on-Insulator Microring Resonator with Bent Couplers", IEEE Group IV Photonics Conference, 08/2015

• Add a model output annotation to show kappa, t and loss at 1550 nm, in the icon for the DC, in INTERCONNECT CML

Doesn't take into account the transformation on the PCell. (offset, rotation, etc).

Add a function to configure environment:

• colour for Marker view
• colour for PCell outline (light blue)
• layer properties
• hotkeys

https://courses.edx.org/courses/course-v1:UBCx+Phot1x+1T2016/discussion/forum/41d69afd67c64e090c77ec3f5fedf2e3b4a025f3/threads/571bad697e622b055a0000f7

Doesn't always snap.

use the closest pins as the alignment feature.

A request to implement a Layout PCell + CML for a fully-parameterized directional coupler with the following parameters:

• g: gap
• r1: Radius 1
• r2: Radius 2
• w: Waveguide width
• Lc: Coupling length
• pol: 'TE' or 'TM'; for use in INTERCONNECT calling FDTD to simulate one of these if pol is defined, or simulate both if undefined.
• wavelength: default 1550. Consider a simulation bandwidth of 10% of the central wavelength. Used by the INTERCONNECT-FDTD simulation.

name: ebeam_dc_seriesrings

generate an icon (SWG file)

See Figure 1 in the following paper:

https://www.osapublishing.org/oe/fulltext.cfm?uri=oe-21-7-9103&id=252045

Once drawn, give user ability to type in a length, and have waveguide adjust itself.

Add Length as parameter in waveguide PCell. Gettable & Settable.

see:
# https://docs.python.org/2/tutorial/floatingpoint.html#tut-fp-issues

and KLayout uses both floating-point and integer storage. So there can be conversion errors.

I created the layout for the ring resonator. My verification step was giving me errors, until I discovered a 1 nm offset in the waveguide due to this type of small error in rounding.